Figure 5.

Comparison of the predicted movement performance on the x- and y-axes and r-coordinates. (a) ${\mathrm{R}}^2$ for the predicted movement performance on the x- (blue dots) and y-axes (green dots) and r-coordinates (data in r-coordinates were calculated as $t_{\mathrm{r},k}=\sqrt{d_{\mathrm{x},k}^2+d_{\mathrm{y},k}^2}$). Each dot indicates ${\mathrm{R}}^2$ in each session. The black dots indicate ${\mathrm{R}}^2$ for the predicted movement performance in r-coordinates and that in r-coordinates by calculating $y_{\mathrm{r},k}=\sqrt{y_{\mathrm{x},k}^2+y_{\mathrm{y},k}^2}$ using the predicted performance on the x- and y-axes. (b) Predicted performance in r-coordinate by directly applying ridge regression in the r-coordinate (red line) and indirectly calculating the predicted performance on the x- and y-axes (cyan line). The black dotted line indicates actual data in the r-coordinate. (c) Predictive power of movement performance in the r-coordinate by Gaussian process regression, a nonlinear method. The horizontal axis indicates ${\mathrm{R}}^2$ for the predicted movement performance on the x-axis (blue dots) and the y-axis (green dots) using ridge regression, and the vertical axis indicates that in the r-coordinate using the Gaussian process. The black dots indicate the same values as in (a). (d) Normalized accuracy rate for the prediction of movement performance on the x-axis (blue dots), y-axis (green dots), and r-coordinates by logistic regression, a classification method. The normalized accuracy rate was calculated by subtracting the accuracy rate at the 1st time bin from that at the 12th time bin.